17results about How to "Reduce contribution" patented technology

The invention relates to the technical field of model lightweight processing, and discloses a digital twin model lightweight processing method, and the method comprises the steps: obtaining a digital twin model, constructing a multi-dimensional twin model parameter importance evaluation objective function, and evaluating the digital twin model parameters obtained through training solving; carrying out lightweight processing on the digital twinborn model parameters based on an importance evaluation result to obtain a lightweight digital twinborn model; and carrying out fine tuning on the lightweight digital twinborn model to obtain the lightweight digital twinborn model after fine tuning. Gradient evaluation and channel gain evaluation are carried out according to the gradient of the digital twin model parameters in the training process and the final training result, the digital twin model parameters with high contribution degree to the output result are selected, redundant parameters are filtered, lightweight processing of the digital twin model is achieved, and the training efficiency is improved. And performing fine tuning processing to retain the performance of the original model to obtain the digital twinborn model which reduces the consumption of computing resources and the storage cost.

The invention discloses a neural radiation field optimization method and equipment based on energy loss uncertainty voting. The method comprises the following steps: acquiring a data set required for training and testing a neural radiation field; an NLL function used by ActiveNeRF training is changed into an energy loss function, and a more reasonable regular term is used; training the model by using the energy loss; building a multi-head attention network model; acquiring a data set for training the multi-head attention network model; training a multi-head attention network model; and the trained attention network model is used to replace the color value of the sampling point with large noise under the test pose, and the final rendering effect is improved. According to the method, data uncertainty is modeled through the Bayesian method, training is carried out through energy loss, and the problem that negative logarithm likelihood loss can generate overlarge variance regardless of whether a predicted mean value is correct or not, so that model training is damaged is solved. On the basis, the data uncertainty is further fused into a multi-head attention network module, and the final rendering quality is improved.

The invention relates to a super-resolution reconstruction method for a non-parametric local self-adaptive super-resolution video, belongs to the technical field of digital image processing, and is particularly suitable for reconstructing a high-resolution image from a low-resolution video sequence in a complex dynamic motion scene. According to the method, a non-parametric local self-adaptive super-division reconstruction model of a video time-space domain is established through global motion estimation in combination with a local gray difference loss function and a BTV regularization item. By introducing a weight matrix and an adaptive threshold parameter, the influence of a motion estimation error on a reconstruction result is reduced, and a gradient descent method is adopted to carry out iterative optimization. According to the method, a high-resolution image can be effectively reconstructed in a complex motion scene, the reconstruction precision of local details is improved, and the robustness to motion errors is enhanced.

The invention provides a plug abnormity detection method and a detection system, which are used for detecting falling and surface flaws of a plug during operation. The method comprises the following steps: firstly, preprocessing original point cloud data to obtain preprocessed point cloud data, and segmenting the preprocessed point cloud data to obtain header data; and then plug falling detection is carried out, if the plug does not fall, plug surface flaw detection is carried out, and if the plug falls, the steps of data preprocessing and plug falling detection are repeated.

The invention discloses a transaction behavior prediction method and device, a storage medium and a computer program product, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining transaction behavior data according to a behavior prediction request if the behavior prediction request is received; obtaining a target important transaction feature in the historical transaction behavior data, the target important transaction feature being a first important transaction feature screened from the historical transaction behavior data according to a preset linear model, the first important transaction feature is an intersection with a second important transaction feature screened out from historical transaction behavior data according to a preset nonlinear model; and screening important features from the transaction behavior data according to the target important transaction features, and inputting the important features into a pre-trained target behavior prediction model to obtain a prediction result. The accuracy of the user transaction behavior prediction result is improved.

The present invention discloses a cross-device ECG federal privacy classification framework method, system and storage medium, and relates to the field of federated learning technology. The specific steps include the following: the server sends initial model parameters to the client; the client trains the initial model using a training data set to obtain basic type recognition model parameters, and sends them to the server in a differential manner; the server receives the basic type recognition model parameters for clustering to obtain clustering results; the clustering results are weighted averaged to obtain the final update parameters of the server model, and the training of the initial model is completed; the ECG data is recognized by the trained initial model, and the respective ECG data type recognition results are output on each client. The present invention can improve data recognition accuracy while protecting user data privacy during ECG health monitoring.

The present invention discloses a data mining method and system based on self-learning, which collects time series image data of different fresh products at various freshness stages, extracts static features and dynamic features from them, and defines labels by analyzing these features, which associate static features with dynamic features to form a comprehensive description; according to the changing rules of dynamic features, products with similar change patterns are classified into the same similarity group; on this basis, the self-learning mining model is trained using these labels and similarity group data, focusing on improving the recognition ability of similar group products; the mining results are fed back to the dynamic spatiotemporal graph of the system to update the node attributes in the graph, thereby continuously optimizing the recognition algorithm. The present invention can not only effectively overcome the recognition difficulties caused by the change of dynamic features, but also provide a stable recognition benchmark through static features, thereby significantly improving the accuracy and reliability of fresh product recognition at different freshness levels.

The invention discloses a gastrointestinal symptom identification method and a sensing system, and relates to the technical field of health informatics and medical information processing, and the method comprises the steps: collecting excitation response and abdominal sound to calculate an acoustic transfer function, and employing a main formant frequency to invert a quality loading index representing a condensation water film effect; constructing an equalization gain based on the transfer function to carry out spectral shape correction on the abdominal sound time-frequency spectrum; constructing a symptom evidence vector and performing adaptive weighted scaling on the symptom evidence vector by using a quality loading index; and finally calculating the distance between the weighted vector and the symptom mode prototype library and mapping the distance into a prompt probability. According to the method, the problems of acoustic link drift and signal distortion caused by microenvironment humidity change in long-time monitoring are solved, and the accuracy and robustness of symptom identification are improved.

This invention discloses an online method and system for estimating the braking clearance throughout the entire lifecycle of an EMB system, relating to the field of braking technology for EMB systems in autonomous vehicles. The invention includes: acquiring motor current and motor angle signals of the electromechanical braking system during the braking clamping phase; constructing a current-angle regression model based on a Legendre orthogonal basis based on the nonlinear relationship between the motor current and motor angle signals; performing online robust estimation of the parameter vector of the current-angle regression model using a weighted multi-forgetting factor recursive least squares algorithm combined with an adaptive M-estimator and a two-stage estimation strategy; and mapping the parameter estimates to braking clearance estimates. This invention achieves high-precision braking clearance estimation without additional sensors based solely on motor current and angle signals through online self-learning. It maintains excellent robustness and stability under complex operating conditions and long-term operation, making it suitable for engineering applications in EMB systems.

The invention relates to the technical field of data processing, in particular to a state prediction method and system for power terminal equipment. The method comprises the steps of collecting output current data and output voltage data, and obtaining a first weight coefficient of each output current data according to an inverse proportion relation between the output current data and the output voltage data; acquiring a second weight coefficient of each piece of output current data according to the surrounding data segment of each piece of output voltage data and the first weight coefficient, and acquiring a prediction error of the current output current data according to the second weight coefficient of each piece of output current data; according to the method, the prediction value of the current output current data is obtained according to the prediction error of the current output current data, the abnormal condition of the current output current data is judged according to the prediction value of the current output current data, and then the abnormity of the transformer is judged, so that the accuracy of abnormity detection is improved, and the influence of noise data is avoided.

The invention provides plug anomaly detection equipment. The plug anomaly detection equipment comprises a plug detection system, a cooling system and a data processing system, the cooling system is used for cooling the plug detection system and the plug working environment; the plug detection system comprises a visual unit used for shooting plug images; the visual unit is in communication connection with the data processing system; and the data processing system comprises a data preprocessing unit, a plug falling detection unit, a plug surface flaw detection unit and an alarm unit, and is used for data acquisition, data processing, plug falling and surface flaw detection when the plug works. The method comprises the following steps: preprocessing original point cloud data to obtain preprocessed point cloud data, and segmenting the preprocessed point cloud data to obtain header data; and then plug falling detection is carried out, if the plug does not fall, plug surface flaw detection is carried out, and if the plug falls, the steps of data preprocessing and plug falling detection are repeated.

This disclosure describes a sensing capability resistant to crosstalk. A current sensor is provided for a target conductor among a plurality of conductors. The current sensor includes at least one magnetic sensor configured to provide two signals representing two different parameters of a field, the two different parameters being different components or their directional derivatives (e.g., gradients). The current sensor also includes a processor configured to derive a signal indicating a current based on a linear combination of a first signal and at least a second signal. At least one of these signals is weighted by coefficients that are constants selected based on the distance between the sensor and at least one of the plurality of conductors in at least a first or second direction. The coefficients are selected to reduce the contribution of parasitic magnetic fields to the signal indicating a current in the first conductor, wherein the parasitic magnetic field is generated by at least one other conductor.

The application discloses a kind of based on LoRa's sea cable state remote wireless telemetry system and method, belong to cable laying state monitoring technical field, system includes being laid on several state monitoring nodes of sea cable, relay node and shipborne monitoring center.Adaptive different compensation strategies are taken according to the proportion of missing packet loss monitored by analysis computer: when mild packet loss, trend filling is executed using effective node change rate and spatial correlation;When moderate packet loss, confidence weight that attenuates dynamically with interruption time is introduced in reconstruction model;When severe packet loss, spatial physical correlation constraint matrix feature sub-block is extracted to perform order reduction calculation and combine spatial interpolation method to recover data.The application solves the problem of data loss caused by unstable wireless communication in extreme sea conditions through hierarchical compensation mechanism, ensuring that the full-line state distribution of sea cable can be reconstructed stably and accurately and the warning can be output under different packet loss conditions, significantly improving the robustness and construction safety of the sea cable monitoring system.

The present invention provides an image classification method and system based on label noise domain adaptation, which respectively obtains a source domain data set composed of image data samples in a source domain and a target domain data set composed of image data samples in a target domain, wherein the image data samples in the source domain contain label noise; constructs an image classification model, performs adaptive noise detection on the image data samples in the source domain and obtains sample weights, divides the source domain data set into a noise sample set and a clean sample set by using a threshold method, performs adaptive noise correction on the clean sample set, calculates the contribution weights of the image data samples to the category domain in the process of aligning the source domain and the target domain tasks, obtains the total model loss from the obtained contribution weights and the sample weights, and obtains a robust noise domain adaptation model through training; obtains the image classification result of the target domain from the obtained robust noise domain adaptation model; the present invention can reduce the influence of label noise on the task alignment of the source domain and the target domain, and effectively improves the accuracy of image classification.

A multimodal brain disease diagnostic system based on phenotypic priors and dual-spectral-domain synergistic enhancement. This system belongs to the interdisciplinary field of artificial intelligence and brain science. It addresses the technical problem that existing methods struggle to meet practical needs in terms of diagnostic accuracy and model robustness when dealing with highly heterogeneous multicenter clinical data. The system described in this invention preserves the spectral heterogeneity of rs-fMRI signals to enhance feature expression; constructs a more reliable population graph through diagnostic conflict suppression; aligns cross-modal representations using contrastive learning and gating mechanisms; and captures multi-scale population structure by combining spectral-domain graph filtering, thereby significantly improving the accuracy, robustness, and interpretability of brain disease diagnosis.

This invention discloses a method for low-altitude UAV target identification and early warning based on multi-source perception, relating to the field of low-altitude security and control technology. The method includes: acquiring low-altitude perception data from radar, photoelectric, and passive radio frequency sensors, and preprocessing it to form standardized multi-source observation data; establishing a source quality assessment model to calculate source quality factors and performing spatiotemporal alignment of the data; outputting the target fusion trajectory state quantity and confidence level through gated probabilistic correlation and adaptive weighted fusion, and extracting multi-dimensional features such as radar micro-Doppler, photoelectric shape, and radio frequency fingerprint to construct a fusion feature vector input classifier to identify target categories; establishing a threat scoring function by combining trajectory prediction and behavioral pattern analysis, outputting threat level and confidence level, and executing graded early warning based on dynamic thresholds. This invention solves the problems of unstable detection by a single sensor, low identification accuracy, and delayed early warning in low-altitude environments, improving the robustness and predictability of low-altitude UAV target control.

A method for predicting coal spontaneous combustion temperature based on adaptive fusion of multi-source heterogeneous data is proposed. This method collects four data sources: time-series data of indicator gas concentrations, distributed fiber optic temperature field data, infrared thermal imaging data, and environmental parameter data. Modal features are extracted using bidirectional long short-term memory networks, one-dimensional convolutional networks, two-dimensional convolutional networks, and fully connected networks, respectively. Data quality scores are calculated in real-time for each data source, generating adaptive fusion weights. A cross-modal multi-head attention mechanism is used to deeply fuse the multi-modal features. The fused features are then propagated multiple times forward using a Monte Carlo random deactivation method, outputting the predicted temperature value and its confidence interval. A risk assessment index is constructed by combining the temperature rise rate, enabling graded early warning of coal spontaneous combustion temperature. This invention overcomes the shortcomings of existing methods, such as reliance on a single data source, lack of sensor fault tolerance, and lack of confidence assessment for prediction results, thus improving the accuracy, robustness, and scientific validity of coal spontaneous combustion temperature prediction and early warning.